Topinformationen

HybrInt – Hybrid Intelligence through Interpretable AI

The aim of this project is to strengthen basic AI research jointly at Leibniz University Hannover and Osnabrück University in the name of hybrid intelligence. The key idea is to combine the strengths of the complementary heterogeneous intelligence of humans and machine: human intelligence is defined by the ability to learn, reason, and interact with the environment based on their knowledge, whereas AI is attributed to machines. This includes tasks, such as language processing, object recognition, model building, and applying that knowledge to solve problems.

With the overarching goal, the research is structured by seven subprojects (SP) ranging from basic research to application and hardware. These project subsume research on knowledge graphs, explainable AI, resource-efficient hardware acceleration, reinforcement learning, trustworthiness, robotics, and human-centered explanations.

• SP1: Knowledge Graph-based Extraction of Research Knowledge from Articles;
• SP2: Knowledge-Graph-Based Reinforcement Learning;
• SP3: Knowledge Graph-based Interpretable Learning on Complex Data;
• SP4: Algorithm Hardware Codesign for Resource-Efficient Interpretable AI Methods;
• SP5: Robust Online Single Plant Classification from Multimodal Sensor Data Including Semantic Context Knowledge;
• SP6: Credible and Structured Interpretations of Machine Learning Models;
• SP7: Human-Centered Explanation of Machine Learning Results.

The envisaged research cooperation should manifest itself in two joint use cases in the field of high precision farming.

1. Optimizing Irrigation – Agricultural Water Management. We aim to develop novel methods, where expert knowledge on irrigation optimization can be incorporated in a human-understandable fashion and new knowledge can be extracted from the learning agent’s experience to enrich human expertise.
2. Agricultural Research Knowledge Observatory. We aim to retrieve relevant literature addressing biodiversity, agricultural, and plant-related knowledge questions, and to create structured contribution descriptions for each of the found articles. In addition, we also aim to link the literature to relevant datasets and possibly other artifacts such as images, videos etc.

HybrInt Retreat - September 13, 2024

To strengthen the cooperation among sub-projects, enhance the work, and discussion within the project as a whole, the Retreat of the HybrInt project has been held at Coppenrath Innovation Centre (CIC) Osnabrück and Gut Arenshorst Bohmte on September 12th and 13th, 2024.

Publications

SP1:

1. Zhiyin Tan, Jennifer D’Souza
   "Bridging the Evaluation Gap: Leveraging Large Language Models for Topic Model Evaluation"
   IRCDL'25: 21st conference on Information and Research science Connecting to Digital and Library science, Feb 20-21, 2025, Udine, Italy.
   Paper: https://ceur-ws.org/Vol-3937/paper15.pdf
   Github: https://github.com/zhiyintan/topic-model-LLMjudgment

SP2:

1. Maximilian Schier, Frederik Schubert, Bodo Rosenhahn
   "Explainable Reinforcement Learning via Dynamic Mixture Policies"
   2025 IEEE International Conference on Robotics and Automation (ICRA), IEEE.
   Paper : https://www.tnt.uni-hannover.de/papers/data/1769/ICRA_2025-4.pdf
   Github : https://github.com/m-schier/Explainable-RL-Dynamic-Mixture-Policies

2. Yannik Mahlau, Frederik Schubert, Bodo Rosenhahn
   "Mastering Zero-Shot Interactions in Cooperative and Competitive Simultaneous Games"
   Proceedings of the 41st International Conference on Machine Learning (ICML), July 2024.
   Paper: https://openreview.net/forum?id=SoqxSnEUi1
   Github:https://github.com/ymahlau/albatross

SP3:

1. Thomas Norrenbrock, Timo Kaiser, Sovan Biswas, Ramesh Manuvinakurike, Bodo Rosenhahn
   "QPM: Discrete Optimization for Globally Interpretable Image Classification"
   The Thirteenth International Conference on Learning Representations (ICLR), April 2025. 
   Paper: openreview.net/forum
   Github: https://github.com/ThomasNorr/QPM

2. Thomas Norrenbrock, Marco Rudolph, Bodo Rosenhahn
   "Q-SENN: Quantized Self-Explaining Neural Networks"
   AAAI Technical Track on Safe, Robust and Responsible AI, AAAI Press, Vol. 38, No. 19, pp. 21482-21491, Vancouver, Canada, February 2024, edited by Michael J. Wooldridge and Jennifer G. Dy and Sriraam Natarajan.
   Paper: https://dl.acm.org/doi/10.1609/aaai.v38i19.30145
   Github: https://github.com/ThomasNorr/Q-SENN

3. Dan Hudson, Jurgen Van Den Hoogen, Stefan Bloemheuvel, Martin Atzmueller
   "Stay tuned! Analysing hyperparameters of a wide-kernel architecture for industrial faults" 
   2024 IEEE Conference on Artificial Intelligence (CAI).
   Paper: https://ieeexplore.ieee.org/document/10605520

4. Jurgen van den Hoogen, Dan Hudson, Martin Atzmueller
   "Graph Signal Processing Unearths the Best Locations for Soil Moisture Sensors"
   2024 International Conference on Machine Learning and Applications (ICMLA).
   Paper: ieeexplore.ieee.org/abstract/document/10903245

SP4:

SP5:

1. Mariia Khan, Yue Qiu, Yuren Cong, Bodo Rosenhahn, David Suter, Jumana Abu-Khalaf
   "Indoor Scene Change Understanding (SCU): Segment, Describe, and Revert Any Change"
   IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , IEEE, Abu Dhabi, United Arab Emirates , October 2024.
   Paper : https://ieeexplore.ieee.org/document/10801354
   Github: https://github.com/mariiak2021/EmbSCU

2. Mathis Kruse, Marco Rudolph, Dominik Woiwode, Bodo Rosenhahn
   "SplatPose & Detect: Pose-Agnostic 3D Anomaly Detection"
   {IEEE/CVF} Conference on Computer Vision and Pattern Recognition, {CVPR} 2024 - Workshops, IEEE, pp. 3950-3960, June 2024.
   Paper : https://arxiv.org/abs/2404.06832
   Github: https://github.com/m-kruse98/SplatPose

SP6:

1. Alexander Dockhorn, Rudolf Kruse
   "An overview of Reinforcement Learning Algorithms for Causaul discovery"
   Handbook of Artificial Intelligence and Machine Learning in Decision Making. Springer Nature.

SP7: